论文部分内容阅读
以固溶处理温度、时间、冷却方式和时效处理温度、时间为输入层节点参数,以质量增加百分比和疲劳循环次数为输出层节点参数,构建了5×15×12×2的四层神经网络结构模型,用于分析热处理工艺对电阻炉寿命的影响,并进行了试验验证和生产线应用。结果表明,该神经网络预测模型各输出参数的相对训练误差均小于5%,相对预测误差均小于6%;与生产线上的热处理工艺相比,神经网络预测的最优热处理工艺参数可使其在600℃×72 h高温氧化后的质量增加从4.66%减少至0.97%,在600~25℃时的疲劳循环次数增加93.16%,从而明显延长电阻炉的寿命。
Taking the solution temperature, time, cooling method and aging treatment temperature and time as the parameters of the input layer nodes, the parameters of the output layer were calculated based on the percentages of mass increase and fatigue cycles, and a 4 × 5 × 15 × 12 × 4 neural network The structure model is used to analyze the influence of heat treatment process on the life of resistance furnace, and the experimental verification and application of the production line are carried out. The results show that the relative training errors of the output parameters of the neural network prediction model are both less than 5% and the relative prediction errors are less than 6%. Compared with the heat treatment process on the production line, the optimal heat treatment process parameters predicted by the neural network can make The mass increase after high temperature oxidation at 600 ℃ × 72h decreased from 4.66% to 0.97%, and the number of fatigue cycles increased by 93.16% at 600 ~ 25 ℃, which significantly prolonged the life of resistance furnace.